Reuters reported exclusively on Oct 26 that the US Department of Justice has probed into a series of car crashes involving Tesla's self-driving system. The marketing hype over self-driving technology has not died down; however, many suppliers and AI companies have recently voiced different opinions on the future of autonomous cars.
An interesting debate over whether we would still be able to enjoy the pleasure of owning and driving a car was brought to one of the panel talks at the Paris Motor Show last week. A study conducted by Michelin Group and market research company Kantar was cited as saying that the younger generation in Europe now see cars not merely as vehicles but as mobility services.
People buy cars for the image it represents, but that might no longer be the case in the decades to come. Michelin VP Erik Grab said the study showed that young people do not find pleasure in driving or owning a car but rather find pleasure in using mobility services, autonomous or not.
Grab said, "it will be another type of pleasure".
Technology-wise, popular opinion seems to be that autonomy without constraints will not happen anytime soon. Denso CTO has said so; Stellantis CTO has said so. Autonomous cars, instead, could see a more promising future with constraint mobility services and cross-state truck and trailer transportation.
From an AI company's perspective, the reason why governments keep self-driving cars strictly regulated is that before a full autonomy could happen, there will always be a human interference, and to handle partial self-driving and partial human driving in a realistic environment becomes problematic.
It will take decades for the transportation system to go from 100% human driving to 100% pure autonomous driving, so the transitional period is the hardest part.
Taiwan-based driving-AI company OmniEyes said the awareness of danger is something that can't be synthesized by robots. The awareness of danger is an abstract sense that should flash through the human minds within milliseconds as well as a highly complicated mechanism too difficult for AI to learn or mimic.
Liability for fatal crashes related to self-driving will also be hard to determine.
Ned Curic, Stellantis CTO, said unconstraint self-driving is a "human-in-a-system problem". In the aerospace industry, autonomous vehicle was a rudimentary technology that has been under development. However, with better technology, better optics, better systems, better sensors, and better processing capabilities, it is still not the way to solve autonomy, Curic said.
Some argued that the complexity of variables in a human-led transport environment could be more clearly captured and detected by a lidar (light-detection and ranging) system. Lidar is still not a "silver bullet" for autonomous driving, said OmniEyes CEO Ting Chou.
One thing that auto parts manufacturers, including lidar makers, have been saying tirelessly is that there will not be a one-size solution that fits all the problem sets. The automotive sector, the space sector, the agricultural sector, or the military sector - whoever inclined to solve autonomy will have to find a sort of multimodal sensor fusion to get around some of the mobility challenges.
According to Veoneer, a Sweden-headquartered American-Swedish automotive technology provider, lidars for high-volume applications, such as autonomous vehicles, need to fulfill following requirements: low cost, light weight, low power consumption, small package, recyclable, and have enough resolution to support car range.
Tesla's full self-driving (FSD) features have been improving and reaching more and more drivers. The carmaker said earlier this month that it has shipped the FSD beta software to more than 160,000 drivers as of this year. The technology, meanwhile, has also drawn attention from the US Federal government.
According to Reuters on Oct 26, the Justice Department prosecutors in Washington and San Francisco are "examining whether Tesla misled consumers, investors and regulators by making unsupported claims about its driver assistance technology's capabilities".
While turning cars into our personal AI-drivers is unlikely for the foreseeable future, self-driving technology could be more widely adopted by robotaxi or long-hail truck fleets because routing for robotaxis in urban areas and long-haul logistics via well-structured highways are relatively simple and limited.
Although Tesla seems to have successfully built up techies' hope with a free-running self-driving car, mobility services will be the most practical aspects for suppliers, software developers, and car OEMs to consider.
E-mobility could create even more possibilities for the society than self-driving cars. Enriching mobility services provided with electric, intelligent cars, such as shared mobility, will be a solid path to take, and it will require increasing cooperation within the auto industry.
The auto industry will need an eco-systemic approach to develop MaaS. Eric Feunteun, COO of Software Republique at Renault Group, pointed out that "none of the big issue we have in our industry or in the mobility industry can be solved in the silo of the tier-2, tier-1, OEMs, and dealers."
He continued that the big challenge needs new form of cooperation from both directly and indirectly from the value chain. As soon as we agree that we cannot overcome the challenges alone, it leads us naturally to cooperate.